SKILL2STEER: One-Pass Hypernetwork for Mapping Natural-Language Skills to Low-Rank Post-Block Activation Shifts
cs.LG
Submitted: 2026-11-03
Updated: 2026-11-03
Code: https://github.com/rlisml/skill2steer
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Weight Updates as Activation Shifts: A Principled Framework for Steering
- Text-to-LoRA: Instant Transformer Adaption
- Doc-to-LoRA: Learning to Instantly Internalize Contexts
- Learning to Decide, Not to Reason: Parameter-Efficient Decision Operators via Low-Rank Activation Steering
- Latent Harness: Compiling Agent Behaviors into Post-Block Activation Shifts
- SkillsInjector: Dynamic Skill Context Construction for LLM Agents
- SHINE: A Scalable In-Context Hypernetwork for Mapping Context to LoRA in a Single Pass
- SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization
- HyperSteer: Activation Steering at Scale with Hypernetworks
- Steering Language Models With Activation Engineering
- Multilingual E5 Text Embeddings: A Technical Report
- Qwen3 Technical Report
- LatentSkill: From In-Context Textual Skills to In-Weight Latent Skills for LLM Agents
- Skill-to-LoRA: From Using Skills to Learning Behaviors for Token-Efficient LLM Agents
- Parametric Skills
- Representation Engineering: A Top-Down Approach to AI Transparency
Related papers
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis
- (Mis)generalization of Helpful-Only Fine-tuning